---
title: "How to price an AI product: seats, usage or outcomes"
description: "How to price an AI product when inference costs grow with use: seats, usage, outcomes or hybrids, with margin math and real GitHub and Salesforce pricing."
image: https://roasexpert.com/hubfs/roas-expert-visuals/cover-pricing-an-ai-product.png
---

[Skip to content](https://roasexpert.com/blog/pricing-an-ai-product#main-content)

[![ROAS Expert Horizontal Logo](https://roasexpert.com/hs-fs/hubfs/image.png?width=164&height=48&name=image.png)](https://roasexpert.com/)

- Services
  
  Show submenu for Services 
  
    - [Growth Marketing](https://roasexpert.com/growth-marketing)
    - [Agentic SEO](https://roasexpert.com/agentic-seo)
    - [Product Consulting](https://roasexpert.com/product-consulting)
    - [Web Development](https://roasexpert.com/web-development)
- Who we help
  
  Show submenu for Who we help 
  
    - [DTC and ecommerce](https://roasexpert.com/dtc-ecommerce)
    - [SaaS](https://roasexpert.com/saas)
    - [AI startups](https://roasexpert.com/ai-startups)
- [Free SEO audit](https://tools.roasexpert.com/?utm_source=roasexpert&utm_medium=nav)
- [Pricing](https://roasexpert.com/pricing)
- [About](https://roasexpert.com/about)
- [Insights](https://roasexpert.com/blog)

Open main navigation

Close main navigation

- Services
  
  Show submenu for Services 
  
    - [Growth Marketing](https://roasexpert.com/growth-marketing)
    - [Agentic SEO](https://roasexpert.com/agentic-seo)
    - [Product Consulting](https://roasexpert.com/product-consulting)
    - [Web Development](https://roasexpert.com/web-development)
- Who we help
  
  Show submenu for Who we help 
  
    - [DTC and ecommerce](https://roasexpert.com/dtc-ecommerce)
    - [SaaS](https://roasexpert.com/saas)
    - [AI startups](https://roasexpert.com/ai-startups)
- [Free SEO audit](https://tools.roasexpert.com/?utm_source=roasexpert&utm_medium=nav)
- [Pricing](https://roasexpert.com/pricing)
- [About](https://roasexpert.com/about)
- [Insights](https://roasexpert.com/blog)
- [Book a call](https://roasexpert.com/contact)

[Book a call](https://roasexpert.com/contact)

 October 4, 2026

# Pricing an AI product: seats, usage or outcomes

![Picture of Rishabh Gupta](https://roasexpert.com/hs-fs/hubfs/IMG_1011.jpeg?width=50&name=IMG_1011.jpeg) [Rishabh Gupta](https://roasexpert.com/blog/author/rishabh-gupta)

Share: [facebook-f icon](http://www.facebook.com/share.php?u=https://roasexpert.com/blog/pricing-an-ai-product) [linkedin-in icon](http://www.linkedin.com/shareArticle?mini=true&url=https://roasexpert.com/blog/pricing-an-ai-product) [Twitter icon](https://twitter.com/intent/tweet?url=https://roasexpert.com/blog/pricing-an-ai-product) [pinterest-p icon](http://pinterest.com/pin/create/link/?url=https://roasexpert.com/blog/pricing-an-ai-product) [envelope icon](mailto:?body=https://roasexpert.com/blog/pricing-an-ai-product)

For most AI products, the right answer is a hybrid: a seat or platform fee that includes a set amount of usage, with a published overage rate after that. Use pure seats only when the AI assists a person and usage per person is fairly even. Use pure usage for technical buyers who can model their volume, and charge per outcome only when you can verify the result and it’s mostly your doing.

The market is already moving this way. GitHub [moved Copilot to usage-based billing](https://github.blog/changelog/2026-06-01-updates-to-github-copilot-billing-and-plans/) with AI credits on June 1, 2026. Salesforce sells Agentforce per user, per conversation and per action. Intercom, Zendesk and HubSpot bill their AI support agents on outcomes such as resolved conversations. ICONIQ’s [2026 State of AI report](https://www.iconiq.com/growth/reports/state-of-ai-2026), based on a Q2 2026 survey of about 300 executives at software companies building AI products, says consumption-based pricing rose from 35% to 42% in six months and outcome-based pricing from 18% to 23%. The average company blends 1.7 pricing models.

In short

- For most AI products the answer is a hybrid: a seat or platform fee with included usage, then a published overage rate
- Pure seats only when the AI assists a person and usage per person is even. Pure usage for technical buyers. Outcomes only when you can verify the result
- Price from the usage spread, not the average: median, 90th percentile and the top 1% by cost
- Set the overage rate from cost: cost per action ÷ (1 − target margin). Your meter should move when your costs do

**50 to 60%**gross margins at AI companies, against 80 to 90% for classic SaaSBessemer, February 2026

**−50%**gross margin on a heavy user at a flat $40 seat: 1,200 actions at $0.05. The same user on a 500-action plan is at 45%Illustration

**42%**of software companies building AI products now use consumption pricing, up from 35% in six monthsICONIQ, State of AI 2026

On this page

1. [Why per-seat pricing breaks for AI products](https://roasexpert.com/blog/pricing-an-ai-product#why-per-seat-pricing-breaks-for-ai-products)
2. [Seats, usage, outcomes and hybrids compared](https://roasexpert.com/blog/pricing-an-ai-product#seats-usage-outcomes-and-hybrids-compared)
3. [Gross margin math: an illustration](https://roasexpert.com/blog/pricing-an-ai-product#gross-margin-math-an-illustration)
4. [How do you choose a value metric?](https://roasexpert.com/blog/pricing-an-ai-product#how-do-you-choose-a-value-metric)
5. [Packaging moves that protect margin](https://roasexpert.com/blog/pricing-an-ai-product#packaging-moves-that-protect-margin)
6. [How do you change pricing without losing trust?](https://roasexpert.com/blog/pricing-an-ai-product#how-do-you-change-pricing-without-losing-trust)
7. [How to test pricing before you set limits](https://roasexpert.com/blog/pricing-an-ai-product#how-to-test-pricing-before-you-set-limits)

## Why per-seat pricing breaks for AI products

Classic SaaS could sell seats because serving one more user cost almost nothing. AI doesn’t work that way. Bessemer’s [AI pricing and monetization playbook](https://www.bvp.com/atlas/the-ai-pricing-and-monetization-playbook) (February 2026) points out that every AI query carries a real cost, and puts AI companies’ gross margins at 50 to 60%, against 80 to 90% for SaaS. Three things go wrong under a flat seat price.

*Problem 01***Cost per user varies**

Your bill scales with actions, tokens and the model behind each request, not with headcount. When GitHub [announced its switch](https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/), it wrote that “a quick chat question and a multi-hour autonomous coding session can cost the user the same amount,” and called its premium request model “no longer sustainable.”

*Problem 02***Heavy users dominate cost**

Explaining its own 2025 change, [Cursor wrote](https://cursor.com/blog/june-2025-pricing) that “the hardest requests cost an order of magnitude more than simple ones.” Under a flat seat, your light users pay for your heavy ones.

*Problem 03***Agents shrink the seat count**

Say your agent closes a third of a customer’s support tickets. At renewal they need fewer support seats, so your revenue falls as your product gets better. Some vendors now price people and agents separately. Salesforce prices employee-facing Agentforce at $125 a user a month with unmetered use, but bills customer-facing agents at $2 a conversation, according to its [pricing page](https://www.salesforce.com/agentforce/pricing/).

## Seats, usage, outcomes and hybrids compared

Here’s how each model bills, with a public example checked on the vendor’s own page in October 2026.

![Three ways to price an AI product: seats are simple to buy but margin drops with heavy users, usage tracks cost but is hard to forecast, outcomes are easy to justify but need clean attribution.](https://roasexpert.com/hubfs/roas-expert-visuals/ai-startups-pricing-models.svg?v=3)

Each model has one strength and one risk. Most AI products land on a hybrid.

| Model and public example | How it bills | When it fits | Main risk |
| --- | --- | --- | --- |
| Per seat: [Microsoft 365 Copilot](https://www.microsoft.com/en-us/microsoft-365/copilot/enterprise) | $30 a user a month, paid yearly | AI assists a person and use per person is even | Heavy users erase margin, and agents shrink seat counts |
| Seats with usage limits: [GitHub Copilot Business](https://docs.github.com/en/copilot/get-started/plans) | $19 a user a month including 1,900 AI credits, then $0.01 a credit | Use varies widely between people on the same team | Users hit limits mid-task |
| Usage: [Salesforce Flex Credits](https://www.salesforce.com/agentforce/pricing/) | $500 per 100,000 credits, with 20 credits ($0.10) per action | Technical buyers, APIs and background agents | Bills are hard to forecast, so buyers ration use |
| Fee with included usage and overage: [Cursor](https://cursor.com/docs/account/pricing) | $20 a month with a set amount of model usage, then on-demand usage at API rates | A mix of light and heavy users | Bill shock if users can’t see the meter |
| Outcome: [Intercom Fin](https://fin.ai/pricing) | $0.99 per outcome, charged at most once per conversation | Results you can verify and mostly control | Disputes over what counts, and you pay for failed attempts |

Microsoft shows a clean split between people and agents. [Microsoft 365 Copilot](https://www.microsoft.com/en-us/microsoft-365/copilot/enterprise) is a $30 seat, but agent use in Copilot Chat can be [billed pay-as-you-go at $0.01 per message](https://learn.microsoft.com/en-us/microsoft-365/copilot/pay-as-you-go/meters) to an Azure subscription. Seats for people, meters for agents is a pattern worth copying.

[GitHub’s plans](https://docs.github.com/en/copilot/get-started/plans) and [Cursor’s plans](https://cursor.com/docs/account/pricing) show the hybrid I’d start most AI apps on: a subscription that includes usage, with pay-as-you-go after it. Salesforce’s [Flex Credits](https://www.salesforce.com/news/press-releases/2025/05/15/agentforce-flexible-pricing-news/), announced in May 2025, meter each agent action. On outcomes, Intercom’s [Fin](https://fin.ai/pricing) charges per outcome, and HubSpot [moved its Customer Agent](https://www.hubspot.com/company-news/hubspots-customer-agent-and-prospecting-agent-now-you-pay-when-the-task-is-complete) from $1.00 per conversation to $0.50 per resolved conversation on April 14, 2026. Notice that HubSpot bills that outcome in credits. Outcome as the trigger and credits as the currency is a hybrid too.

## Gross margin math: an illustration

Gross margin per user is price minus inference cost, divided by price. Inference cost is actions times your cost per action. Run it across your real usage spread, not for the average user.

Here’s an illustration with made-up numbers. Say you sell an AI assistant at $40 a seat a month, and each action, such as a research task or a drafted email, costs you $0.05 in inference.

| User | Actions a month | Inference cost | Gross margin |
| --- | --- | --- | --- |
| Light | 100 | $5 | 87.5% |
| Typical | 300 | $15 | 62.5% |
| Heavy | 1,200 | $60 | -50% |

Past 800 actions a month, a seat loses money. Now spread 100 seats as 50 light, 40 typical and 10 heavy. Revenue is $4,000, inference is $1,450 and blended margin is 63.75%. That looks fine until you see that the heavy 10% of seats drive 41% of inference cost. Your margin now depends on who signs up next.

**Inference cost per user against a flat $40 seat**

 $0$20$40$60Light: 100 actions (87.5% margin) Light: 100 actions (87.5% margin): $5 $5Typical: 300 actions (62.5% margin) Typical: 300 actions (62.5% margin): $15 $15Heavy: 1,200 actions (−50% margin) Heavy: 1,200 actions (−50% margin): $60 $60

 $0$20$40$60Light: 100 actions (87.5% margin) Light: 100 actions (87.5% margin): $5 $5Typical: 300 actions (62.5% margin) Typical: 300 actions (62.5% margin): $15 $15Heavy: 1,200 actions (−50% margin) Heavy: 1,200 actions (−50% margin): $60 $60

At $0.05 per action, the seat breaks even at 800 actions. Spread 100 seats as 50 light, 40 typical and 10 heavy and blended margin is 63.75%, with the heavy 10% driving 41% of inference cost.

Add a limit. Keep the $40 seat, include 500 actions and charge $0.10 for each action after that. Light and typical users never notice. The heavy user pays $110 against $60 of cost, a 45% margin instead of minus 50%, and blended margin rises to about 69%.

**The heavy user: what they pay against what they cost**

 $0$40$80$120Flat seat: pays Flat seat: pays: $40 $40Flat seat: costs Flat seat: costs: $60 $60500 included + $0.10 overage: pays 500 included + $0.10 overage: pays: $110 $110500 included + $0.10 overage: costs 500 included + $0.10 overage: costs: $60 $60Same plan, cost per action doubled: pays Same plan, cost per action doubled: pays: $110 $110Same plan, cost per action doubled: costs Same plan, cost per action doubled: costs: $120 $120

 $0$40$80$120Flat seat: pays Flat seat: pays: $40 $40Flat seat: costs Flat seat: costs: $60 $60500 included + $0.10 overage: pays 500 included + $0.10 overage: pays: $110 $110500 included + $0.10 overage: costs 500 included + $0.10 overage: costs: $60 $60Same plan, cost per action doubled: pays Same plan, cost per action doubled: pays: $110 $110Same plan, cost per action doubled: costs Same plan, cost per action doubled: costs: $120 $120

Red bars are revenue, grey bars cost. The limit turns a −50% margin into 45% and lifts blended margin to about 69%. Double the cost per action and a $0.10 overage only matches cost, so set it from cost: $0.125 at $0.05 and a 60% target.

Then stress test the model. If you switch to a model that doubles your cost per action to $0.10, the typical flat-seat user falls to a 25% margin and the heavy one costs you $120 against $40 of revenue. The 500-action plan breaks too. The heavy user pays $110 and costs $120, because a $0.10 overage now only matches cost, and blended margin drops from about 69% to 38%. Set the overage rate from cost instead: divide cost per action by one minus your target margin. At $0.05 and a 60% target, that’s $0.125. GitHub now [converts tokens into credits](https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/) at each model’s published API rate, so its meter tracks model cost. Your meter should move when your costs do.

Outcome pricing changes the denominator. You pay inference on every attempt but bill only the wins. At $1.00 per resolution and $0.15 of inference per conversation, a 60% resolution rate means $0.25 of cost per billed resolution, a 75% margin. At a 30% resolution rate, cost per billed resolution doubles to $0.50 and margin drops to 50%. A customer with a thin help center can double your cost per resolution.

This is the same discipline we apply to paid media: judge spend on profit, not revenue. If you buy growth with ads, this margin drives the payback math in [ROAS, MER and POAS](https://roasexpert.com/blog/roas-mer-poas) and the bid caps in [paid acquisition for AI apps](https://roasexpert.com/blog/paid-acquisition-for-ai-apps).

**Outcome pricing: cost per billed resolution**

 $0.00$0.25$0.5060% resolution rate 60% resolution rate: $0.25 $0.2530% resolution rate 30% resolution rate: $0.50 $0.50

 $0.00$0.25$0.5060% resolution rate 60% resolution rate: $0.25 $0.2530% resolution rate 30% resolution rate: $0.50 $0.50

At $1.00 per resolution and $0.15 of inference per conversation. You pay inference on every attempt and bill only the wins, so a customer with a thin help center can halve your margin.

## How do you choose a value metric?

The value metric is the unit your price scales with. A good one passes four tests.

*Test 01***It moves with customer value**

When customers get more from the product, they use more of the metric.

*Test 02***The buyer can predict it**

Their finance team can forecast next quarter’s bill without calling you.

*Test 03***Both sides can measure it**

The customer sees the same count you bill from, as it happens.

*Test 04***It covers your cost**

When your inference bill rises, so does the metric.

Here’s how the common metrics score.

| Value metric | Passes | Fails or depends |
| --- | --- | --- |
| Seats | Predictable, measurable | Stops covering cost once usage spreads out |
| Tokens | Measurable, covers cost | Doesn’t move with value, and only engineers can predict it |
| Credits | Measurable, covers cost | Tokens with a friendlier label, so predictable only with included pools and caps |
| Resolutions | Moves with value | Measurable only with a published definition, and covers cost only if your price covers failed attempts |

Published definitions differ, so read them closely. Intercom [counts a Fin resolution](https://fin.ai/help/en/articles/13975800-fin-pricing-outcomes) when the customer confirms it, or when the customer goes quiet for 24 hours after Fin’s last answer. It doesn’t bill when the customer asks for a human. It does bill a second outcome type, the Procedure handoff, when Fin completes a workflow you built to end with a handoff to your team. Zendesk [announced outcome-based pricing](https://www.zendesk.com/newsroom/articles/zendesk-outcome-based-pricing/) in August 2024. Since May 18, 2026, only [verified resolutions](https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers) count against its allowance, and an LLM reviews each conversation to confirm the request was resolved. Pick a definition, write it down and show it on every invoice.

## Packaging moves that protect margin

*Move 01***Included usage**

Size it from percentiles, not the average, because heavy users pull the average up. In the illustration above, the median seat uses 200 actions, the average 290 and the 90th percentile 300, so a 500-action allowance covers 90% of seats in full. GitHub Copilot Business includes 1,900 AI credits per user and [pools them](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-organizations-and-enterprises), so 100 users share 190,000 credits. Pooling is friendlier to buyers, but your margin then lives at the account level. Pool those 500 actions across 100 seats and the 50,000-action allowance covers all 29,000 actions, so overage never bills.

*Move 02***Rollover**

[GitHub](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-organizations-and-enterprises) and [HubSpot](https://www.hubspot.com/products/artificial-intelligence/credits) both expire unused included credits at the end of each month. That stops unused credits piling up into a usage spike later. If you allow rollover, cap it at one month’s allowance.

*Move 03***Overage rates**

Publish them. GitHub [charges $0.01](https://docs.github.com/en/copilot/get-started/plans) per extra AI credit. [Zendesk](https://www.zendesk.com/pricing/) charges $1.50 per extra automated resolution on a commitment and $2.00 pay-as-you-go. HubSpot’s [default](https://knowledge.hubspot.com/account-management/understand-hubspot-credits-and-billing), once you’ve bought extra credits and hit your monthly limit, is to move you to a larger credit pack for the rest of your term. Per-credit overage is a setting you turn on.

*Move 04***Spend caps and alerts**

GitHub’s user-level [budgets](https://docs.github.com/en/copilot/concepts/billing-and-usage/organizations-and-enterprises/budgets) are a hard stop, while its organization and enterprise budgets stop usage only if an admin turns that setting on. HubSpot [alerts admins](https://knowledge.hubspot.com/account-management/understand-hubspot-credits-and-billing) at 75%, 85% and 90% of credits used. Turn caps on by default for new accounts, and send the first alert early enough for the buyer to act.

*Move 05***Annual commits**

Trade a lower unit rate for committed volume. Salesforce sells Flex Credits [pay-as-you-go or pre-purchased](https://www.salesforce.com/news/stories/new-agentforce-payment-options/), and says pre-purchase gets the best overall rate. Commits give you forecastable revenue and give the buyer a number to budget.

## How do you change pricing without losing trust?

Cursor shows the cost of getting this wrong. On June 16, 2025 it replaced the Pro plan’s 500 monthly requests with “$20 of frontier model usage per month at API pricing.” [TechCrunch reported](https://techcrunch.com/2025/07/07/cursor-apologizes-for-unclear-pricing-changes-that-upset-users/) users running out of usage quickly and getting unexpected charges. On July 4, Cursor CEO Michael Truell [wrote](https://cursor.com/blog/june-2025-pricing) that the changes “were not communicated clearly, and we take full responsibility.” Cursor offered refunds for unexpected usage between June 16 and July 4, improved its pricing page and docs, and added usage-limit visibility to its dashboard.

GitHub’s 2026 switch shows the other way. It [announced usage-based billing](https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/) on April 27, five weeks before the June 1 start. Annual Pro and Pro+ subscribers kept premium request pricing until their plans expire. Business and Enterprise customers got larger promotional credit allowances for June, July and August. GitHub promised a preview bill for early May and, on May 12, published [downloadable reports](https://github.blog/changelog/2026-05-12-april-reports-are-now-available-to-prepare-for-usage-based-billing/) showing how each customer’s April activity would translate into credits.

The rules that follow:

1. Give notice measured in billing cycles, and more for annual contracts.
2. Grandfather existing contracts until renewal.
3. Show each customer their own bill under the new price before it applies, and give new buyers a calculator.
4. Refund surprises fast, and say plainly what you got wrong.

## How to test pricing before you set limits

*Step 01***Instrument cost per account first**

Log actions, tokens, model and cost for every user from the first beta. You can’t price a distribution you haven’t measured.

*Step 02***Read the distribution, not the average**

Look at the median, the 90th percentile and the top 1% of users by actions and by cost. Set included usage so the typical user never sees a limit, and price overage for the tail.

*Step 03***Interview the buyer, not just the user**

Ask what the job costs them today in people, tools or agencies, whether they budget for it as a fixed annual line or a variable one, who approves overage, and which result they’d pay for per unit.

*Step 04***Measure willingness to pay**

Van Westendorp’s price sensitivity meter asks four questions: at what price the product is so cheap you’d doubt its quality, a bargain, getting expensive, and too expensive to consider. It gives you a range of acceptable prices, not a single number. [Sawtooth Software](https://sawtoothsoftware.com/resources/blog/posts/van-westendorp-pricing-sensitivity-meter) notes it suits products new to the market and leaves out competitors, so pair it with your interviews and add a purchase-intent question.

*Step 05***Pilot on new customers**

Launch the new package to new signups first, then watch overage, upgrades and churn by cohort for two or three billing cycles before you move existing customers.

## If you want help

Pricing is part of our [product consulting](https://roasexpert.com/product-consulting) work with [AI startups](https://roasexpert.com/ai-startups) and SaaS teams. A product and GTM teardown costs $2,000, and a fractional product lead starts at $3,000 a month for one day a week. [Book a 30-minute call](https://roasexpert.com/contact) to talk it through.

**Pressure-test your pricing before you set limits**

Pricing is part of our product consulting work with AI startups and SaaS teams. A product and GTM teardown costs $2,000, and a fractional product lead starts at $3,000 a month for one day a week.

[Book a 30-minute call](https://roasexpert.com/contact)[Product consulting](https://roasexpert.com/product-consulting)

## Sources

- [State of AI: The Builder’s Economy (2026), ICONIQ](https://www.iconiq.com/growth/reports/state-of-ai-2026)
- [The AI pricing and monetization playbook, Bessemer Venture Partners](https://www.bvp.com/atlas/the-ai-pricing-and-monetization-playbook)
- [GitHub Copilot is moving to usage-based billing, The GitHub Blog](https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/)
- [Updates to GitHub Copilot billing and plans, GitHub Changelog](https://github.blog/changelog/2026-06-01-updates-to-github-copilot-billing-and-plans/)
- [Plans for GitHub Copilot, GitHub Docs](https://docs.github.com/en/copilot/get-started/plans)
- [Usage-based billing for organizations and enterprises, GitHub Docs](https://docs.github.com/copilot/concepts/billing/usage-based-billing-for-organizations-and-enterprises)
- [Budgets for usage-based billing, GitHub Docs](https://docs.github.com/en/copilot/concepts/billing-and-usage/organizations-and-enterprises/budgets)
- [April reports are now available to prepare for usage-based billing, GitHub Changelog](https://github.blog/changelog/2026-05-12-april-reports-are-now-available-to-prepare-for-usage-based-billing/)
- [Clarifying our pricing, Cursor](https://cursor.com/blog/june-2025-pricing)
- [Pricing, Cursor Docs](https://cursor.com/docs/account/pricing)
- [Cursor apologizes for unclear pricing changes that upset users, TechCrunch](https://techcrunch.com/2025/07/07/cursor-apologizes-for-unclear-pricing-changes-that-upset-users/)
- [Agentforce pricing, Salesforce](https://www.salesforce.com/agentforce/pricing/)
- [Salesforce Introduces New Flexible Agentforce Pricing to Accelerate the Digital Labor Revolution, Salesforce](https://www.salesforce.com/news/press-releases/2025/05/15/agentforce-flexible-pricing-news/)
- [New Ways to Pay Make It Easier than Ever to Get Started with Agentforce, Salesforce](https://www.salesforce.com/news/stories/new-agentforce-payment-options/)
- [Microsoft 365 Copilot for enterprise, Microsoft](https://www.microsoft.com/en-us/microsoft-365/copilot/enterprise)
- [Meters for Microsoft Copilot pay-as-you-go services, Microsoft Learn](https://learn.microsoft.com/en-us/microsoft-365/copilot/pay-as-you-go/meters)
- [Fin AI Agent pricing, Intercom](https://fin.ai/pricing)
- [Fin pricing: Outcomes, Fin Help Center](https://fin.ai/help/en/articles/13975800-fin-pricing-outcomes)
- [Zendesk First in CX Industry to offer Outcome-Based Pricing for AI Agents, Zendesk](https://www.zendesk.com/newsroom/articles/zendesk-outcome-based-pricing/)
- [Zendesk pricing plans, Zendesk](https://www.zendesk.com/pricing/)
- [About automated resolution tiers, Zendesk Help](https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers)
- [HubSpot’s Customer Agent and Prospecting Agent: Now you pay when the task is complete, HubSpot](https://www.hubspot.com/company-news/hubspots-customer-agent-and-prospecting-agent-now-you-pay-when-the-task-is-complete)
- [HubSpot Credits, HubSpot](https://www.hubspot.com/products/artificial-intelligence/credits)
- [Manage HubSpot Credits, HubSpot Knowledge Base](https://knowledge.hubspot.com/account-management/understand-hubspot-credits-and-billing)
- [Van Westendorp Pricing Model: Definition, How It Works, Examples, and More, Sawtooth Software](https://sawtoothsoftware.com/resources/blog/posts/van-westendorp-pricing-sensitivity-meter)

Facts checked on 5 October 2026.

ROAS Expert is also listed on [DesignRush Verified Agency](https://www.designrush.com/agency/paid-media-pay-per-click).

[Product Consulting](https://roasexpert.com/blog/tag/product-consulting)

## Related posts

[Growth Marketing](https://roasexpert.com/blog/tag/growth-marketing), [Product Consulting](https://roasexpert.com/blog/tag/product-consulting)

## [Usage-based pricing examples: 8 models explained, with the numbers](https://roasexpert.com/blog/usage-based-pricing-examples)

![Picture of Rishabh Gupta](https://roasexpert.com/hs-fs/hubfs/IMG_1011.jpeg?width=50&name=IMG_1011.jpeg) [Rishabh Gupta](https://roasexpert.com/blog/author/rishabh-gupta) 

 October 20, 2026

By Rishabh Gupta, founder of ROAS Expert. Last updated 20 October 2026.

[Read more](https://roasexpert.com/blog/usage-based-pricing-examples)

[![Cover image for the post: Paid acquisition for AI consumer apps: buy payers, not installs](https://roasexpert.com/hs-fs/hubfs/roas-expert-visuals/cover-paid-acquisition-for-ai-apps.png?height=200&name=cover-paid-acquisition-for-ai-apps.png)](https://roasexpert.com/blog/paid-acquisition-for-ai-apps)

[Growth Marketing](https://roasexpert.com/blog/tag/growth-marketing)

## [Paid acquisition for AI consumer apps: buy payers, not installs](https://roasexpert.com/blog/paid-acquisition-for-ai-apps)

![Picture of Rishabh Gupta](https://roasexpert.com/hs-fs/hubfs/IMG_1011.jpeg?width=50&name=IMG_1011.jpeg) [Rishabh Gupta](https://roasexpert.com/blog/author/rishabh-gupta) 

 October 5, 2026

Buy paying subscribers, not installs. Every active user of an AI app costs you inference, so an...

[Read more](https://roasexpert.com/blog/paid-acquisition-for-ai-apps)

[![Cover image for the post: Product feeds for AI shopping: what ChatGPT and Google actually read](https://roasexpert.com/hs-fs/hubfs/roas-expert-visuals/cover-product-feeds-for-ai-shopping.png?height=200&name=cover-product-feeds-for-ai-shopping.png)](https://roasexpert.com/blog/product-feeds-for-ai-shopping)

[Agentic SEO](https://roasexpert.com/blog/tag/agentic-seo)

## [Product feeds for AI shopping: what ChatGPT and Google actually read](https://roasexpert.com/blog/product-feeds-for-ai-shopping)

![Picture of Rishabh Gupta](https://roasexpert.com/hs-fs/hubfs/IMG_1011.jpeg?width=50&name=IMG_1011.jpeg) [Rishabh Gupta](https://roasexpert.com/blog/author/rishabh-gupta) 

 October 4, 2026

ChatGPT and Google read the same core facts from a product feed: a stable ID, title, description,...

[Read more](https://roasexpert.com/blog/product-feeds-for-ai-shopping)

[![ROAS Expert Logo](https://roasexpert.com/hubfs/image%20(1).svg "ROAS Expert Logo")](http://ROAS%20Expert)

ROAS Expert is a small senior growth team for DTC brands, SaaS and AI startups.

[rishabh@roasexpert.com](mailto:rishabh@roasexpert.com)

- [Pricing](https://roasexpert.com/pricing)
- [About](https://roasexpert.com/about)
- [Insights](https://roasexpert.com/blog)
- [Free SEO audit](https://tools.roasexpert.com/?utm_source=roasexpert&utm_medium=footer)
- [Contact](https://roasexpert.com/contact)
- [Privacy](https://roasexpert.com/legal)
- [Terms](https://roasexpert.com/terms)

---

[Follow us on LinkedIn](https://www.linkedin.com/company/roas-expert/) [Follow us on Instagram](https://instagram.com/roasexperts) [Follow us on Twitter](https://x.com/roasexpert)

© 2026 ROAS Expert, a brand of Westside Techno Consultants, Bilaspur, Chhattisgarh, India.  
Working with brands in the US, UK, EU and the Gulf.  
ROAS Expert is a registered Shopify Partner.

[Contact Us](https://roasexpert.com/contact)

```json
{
  "@context" : "https://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Rishabh Gupta",
    "url" : "https://roasexpert.com/blog/author/rishabh-gupta"
  },
  "dateModified" : "2026-10-21T05:18:56.974Z",
  "datePublished" : "2026-10-04T10:50:02.000Z",
  "headline" : "How to price an AI product: seats, usage or outcomes",
  "image" : [ "https://roasexpert.com/hubfs/roas-expert-visuals/cover-pricing-an-ai-product.png" ],
  "mainEntityOfPage" : {
    "@id" : "https://roasexpert.com/blog/pricing-an-ai-product",
    "@type" : "WebPage"
  },
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "url" : "https://roasexpert.com/hubfs/image%20(1).svg"
    },
    "name" : "ROAS Expert"
  }
}
```

```json
{
  "@context" : "https://schema.org",
  "@id" : "https://roasexpert.com/#org",
  "@type" : "Organization",
  "description" : "Growth team for DTC brands, SaaS and AI startups: growth marketing, agentic SEO, product consulting and web development.",
  "email" : "rishabh@roasexpert.com",
  "founder" : {
    "@id" : "https://roasexpert.com/#rishabh"
  },
  "logo" : "https://roasexpert.com/hubfs/image.png",
  "name" : "ROAS Expert",
  "sameAs" : [ "https://www.linkedin.com/company/roas-expert/", "https://www.crunchbase.com/organization/roas-expert", "https://wellfound.com/company/roas-expert", "https://www.upwork.com/agencies/1631604300620173312/", "https://x.com/roasexpert", "https://www.instagram.com/roasexperts/" ],
  "url" : "https://roasexpert.com/"
}
```

```json
{
  "@context" : "https://schema.org",
  "@id" : "https://roasexpert.com/#website",
  "@type" : "WebSite",
  "alternateName" : "ROAS Experts",
  "name" : "ROAS Expert",
  "publisher" : {
    "@id" : "https://roasexpert.com/#org"
  },
  "url" : "https://roasexpert.com/"
}
```

```json
{
  "@context" : "https://schema.org",
  "@id" : "https://roasexpert.com/#rishabh",
  "@type" : "Person",
  "jobTitle" : "Founder",
  "name" : "Rishabh Gupta",
  "sameAs" : [ "https://www.linkedin.com/in/rishabh16/", "https://www.crunchbase.com/person/rishabh-gupta-fa2e", "https://adplist.org/mentors/rishabh-gupta", "https://mentorcruise.com/mentor/rishabhgupta/", "https://contra.com/rishabhg", "https://www.upwork.com/freelancers/~01f80e0179ab49fa4d", "https://clarity.fm/rishabh16", "https://topmate.io/rishabhg" ],
  "worksFor" : {
    "@id" : "https://roasexpert.com/#org"
  }
}
```